How to increase the efficiency of this MATLAB program by using vectorization techniques

Question

hmhuang 2021-10-27

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1572838-how-to-increase-the-efficiency-of-this-matlab-program-by-using-vectorization-techniques

回答： Rik 2021-10-27

nodetraffic.mat

I have a MATLAB code snippet as below. For now I have tried two versions of implementation to apply a function standard_deviation_distance to each element in an array. I wonder if there are any potential improvements on the code efficiency/performance in terms of the running time optimization (i.e., I hope the code can be run as fast as possible). Any suggestions will be greatly appreciated!

P.S. Please see the file nodetraffic.mat in the attached

load nodetraffic.mat;
% Method I
tic
count1 = 0;
for i = 1 : length(nodetraffic)
    if (standard_deviation_distance(nodetraffic, nodetraffic(i)) > 6)
        count1 = count1 + 1;
    end
end
count1
toc
%% Method II
tic
A = arrayfun(@(x) standard_deviation_distance(nodetraffic, x), nodetraffic);
count2 = length(A(A > 6))
toc
function dist = standard_deviation_distance(v, x)  
    standard_deviation = std(v);
    dist = (x - mean(v)) / standard_deviation;
end

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Rik 2021-10-27

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1572838-how-to-increase-the-efficiency-of-this-matlab-program-by-using-vectorization-techniques#answer_817743

在 MATLAB Online 中打开

nodetraffic.mat

Arrayfun will only hide the loop, so it will never be faster than a plain for loop.

What you need to do is change your function so it allows array inputs. As far as I can tell, there is only a small change required.

fn=websave('data.mat','https://www.mathworks.com/matlabcentral/answers/uploaded_files/780763/nodetraffic.mat');
S=load(fn);
nodetraffic=S.nodetraffic;
method1(nodetraffic)
ans = 24
method2(nodetraffic)
ans = 24
%warm up the run online engine
timeit(@()method1(nodetraffic));timeit(@()method2(nodetraffic));
timeit(@()method1(nodetraffic)),timeit(@()method2(nodetraffic))
ans = 0.2317
ans = 5.2963e-05
function count=method1(nodetraffic)
count = 0;
for i = 1 : length(nodetraffic)
    if (standard_deviation_distance(nodetraffic, nodetraffic(i)) > 6)
        count = count + 1;
    end
end
end
function count=method2(nodetraffic)
d=standard_deviation_distance(nodetraffic, nodetraffic);
count=sum(d>6);
end
function dist = standard_deviation_distance(v, x)
    standard_deviation = std(v);
    dist = (x - mean(v)) ./ standard_deviation;
    %                    ^
    %      only change required here
end